Presenilin 1 mutations activate gamma 42-secretase but reciprocally inhibit epsilon-secretase cleavage of amyloid precursor protein (APP) and S3-cleavage of notch.

The presenilin 1 (PS1) and presenilin 2 (PS2) proteins are necessary for proteolytic cleavage of the amyloid precursor protein (APP) within its transmembrane domain. One of these cleavage events (termed gamma-secretase) generates the C-terminal end of the Abeta-peptide by proteolysis near residue 710 or 712 of APP(770). Another event (termed gamma-like or epsilon-secretase cleavage) cleaves near residue 721 at approximately 2-5 residues inside the cytoplasmic membrane boundary to generate a series of stable, C-terminal APP fragments. This latter cleavage is analogous to S3-cleavage of Notch. We report here that specific mutations in the N terminus, loop, or C terminus of PS1 all increase the production of Abeta(42) but cause inhibition of both epsilon-secretase cleavage of APP and S3-cleavage of Notch. These data support the hypothesis that epsilon-cleavage of APP and S3-cleavage of Notch are similar events. They also argue that, although both the gamma-site and the epsilon-site cleavage of APP are presenilin-dependent, they are likely to be independent catalytic events.

All pathogenic presenilin mutations, which cause familial Alzheimer disease (FAD), enhance the production of the highly amyloidogenic A␤ 42 variant (8 -10). Consequently, it has been suggested that PS1 and PS2 mutations have a "gain-of-function" effect on ␥-secretase activity. Paradoxically, several of these mutations suppress the ability of human presenilin proteins to complement the Notch signaling phenotype induced by the loss-of-function C60S mutant in Caenorhabditis elegans (11,12). Conversely, the C60S loss-of-function mutation in the sel-12 presenilin homologue, which inhibits Notch processing of C. elegans, causes increased A␤ 42 production when introduced into human PS1 as the C92S mutation (13,14). However, the ␥-secretase cleavages of APP near residue 42 (15), the ⑀-cleavages of APP near residue 50 (1)(2)(3)(4)(5), and the S3-cleavage of Notch (7,16) all appear to be presenilin-dependent. It has been difficult, therefore, to reconcile the apparently opposing action of pathogenic presenilin mutations on A␤ production in mammalian cells and on Notch cleavage in invertebrates. We now report that, although PS1 mutations augment ␥-secretase cleavage at residue 42, they in fact suppress both S3-cleavage of Notch and ⑀-cleavage of APP near residue 50. These data suggest that ␥-secretase cleavage at residue 42 and S3-/⑀-cleavage are likely to be independent activities.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfection-Human embryonic kidney 293 (HEK293) cell lines were stably transfected with wild-type PS1 or various mutant PS1s including L392V, G206A, Exon 9 missplicing mutation (⌬Exon 9), or D385A. Certain cell lines also stably expressed the APP 695 "Swedish" mutant. In some experiments, cells were transiently transfected with cDNA encoding a membrane-bound Notch protein (Myc-tagged Notch⌬E) cloned into pCS2ϩ vector (17,18). To examine the ability of APP mutants to increase or decrease A␤ production, some cells were co-transfected with cDNAs encoding wild-type or mutant forms of the C-terminal 99 amino acids of APP (corresponding to the product of ␤-secretase cleavage of FL-APP) that had been fused to an artificial signal peptide to ensure correct orientation and cloned into pCEP4 vector (19). Cell lines were maintained in 10% fetal bovine serum Dulbecco's modified Eagle's medium with 200 g/ml Zeocin (Invitrogen) or/and 10 g/ml blasticidin (Invitrogen) or/and 200 g/ml G418 (Invitrogen).
Generation of ⑀-Stubs and A␤ in Cell-free Assay Systems-The generation of ⑀-stubs and A␤ was performed in cell-free assay systems as described previously (4,20). HEK293 cells were resuspended in 5 mM HEPES, pH 7.4, 1 mM EDTA, 0.25 M sucrose, plus protease inhibitors and homogenized, and the postnuclear supernatant was prepared as described (21). Microsomal membranes pelleted from the postnuclear supernatant by centrifugation at 100,000 ϫ g for 1 h at 4°C or detergent-solubilized, purified, high molecular weight PS1 complexes (see below) were washed once with assay buffer and resuspended in assay buffer, pH 6.4 (10 mM KOAc, 1.5 mM MgCl 2 , and 75 mM sodium citrate) containing protease inhibitors and were then incubated at 37°C for the indicated times to generate ⑀-stubs and A␤. Control samples were kept on ice. At the end of the assay, the microsomal membrane samples were either used for Western blotting or separated into pellet and supernatant fractions by ultracentrifugation for 1 h at 100,000 ϫ g at 4°C. The purified PS1 complexes were prepared for SDS-PAGE by adding 2ϫ sample buffer.
Western Blot and Immunoprecipitation-Purified, detergent-solubilized, high molecular weight PS1 complexes or crude microsomal membrane preparations were lysed by adding 1% Nonidet P-40 in Trisbuffered saline. After protein assay, equal amounts of total protein from each sample were dissolved in SDS sample buffer, separated on SDS-PAGE, and transferred to nitrocellulose membrane. The target proteins were visualized by enhanced chemiluminescence (Amersham Biosciences) with antibodies to the C terminus of APP (APP C-terminal polyclonal antibody, Sigma), to PS1-NTF (antibody 14, gift from Dr. S.E. Gandy), and to A␤ (3340 and 3542, gifts from Dr. F. Checler; 6E10 and 4G8, Signet Laboratories, Inc.). Where necessary, nitrocellulose membranes were stripped for reuse. A␤ immunoprecipitation from the reaction lysates (A␤ generated in vitro from cell-free preparations) or from conditioned medium (A␤ secreted from intact cells) was performed as described previously (17) using A␤ antibodies (6E10, 4G8, 3340, or 3542). For detection and electrophoretic separation of A␤ 40 and A␤ 42 , the Bicine/Tris SDS-PAGE system was employed as described (22).
Metabolic Labeling and Quantification of NICD and AICD-For pulse-chase experiments, cells were metabolically labeled with [ 35 S]methionine/cysteine (ICN Pharmaceuticals) for 20 min and chased for 1 h (maximum NICD formation occurred at ϳ1 h) (17,21,23). Cell lysates were subjected to immunoprecipitation with an antibody to the myc epitope fused to the C terminus of Notch⌬E.
The intensities of the autoradiographic bands or ECL signals for Notch, APP-CTFs (including ⑀-stub/AICD), and FL-APP were measured by both densitometry and/or PhosphorImager analysis. The ratios of ⑀-stub to C99-APP or FL-APP and of NICD to total Notch protein (NICD ϩ Notch⌬E) were calculated for each cell line. To permit comparison across different replications, we normalized these ratios to those in cells expressing wild-type PS1, which was set at 1.00. Pairwise comparisons were made using Student's t test for independent groups for the preplanned comparisons of ⑀-stub and NICD generation normalized to their production in cells expressing PS1-wt.
Mass Spectrometric Analysis of APP C-terminal Stubs-The products of ␥or ⑀-cleavage were immunoprecipitated from the supernatant fractions from either the crude microsomal membrane preparations or from purified PS1 complexes using anti-APP-CTF monoclonal antibody C1/6.1 (gift from Paul Mathews). The captured C-terminal APP stubs were eluted from beads with 5 l of 1% trifluoroacetic acid/30% acetonitrile. To prepare the sample matrix solution, 1 l of immunoprecipitated ⑀-stub sample was mixed with 1 l of saturated ␣-cyano-4-hydroxycinnamic acid solution in 0.1% trifluoroacetic acid/acetonitrile. The sample matrix solution was then applied to the sample plate and dried at ambient temperature prior to mass spectrometric analysis. Matrix-assisted laser desorption/ionization mass spectrometric (MALDI-MS) analysis was performed on a Voyager-DE STR mass spectrometer.
Protein Assay-Protein concentration was determined using the BCA protein assay reagents according to the manufacturer's instructions.
Replications-All experiments were performed at least in triplicate.

Generation of APP C-terminal Stubs by Microsomal Membranes and by Purified PS1
Complexes-Incubation of microsomal membrane fractions derived either from native HEK293 cells or from HEK293 cells expressing APP Swedish generated an ϳ6-kDa APP C-terminal fragment, which migrated below the major APP C-terminal fragments (i.e. ␣-stub and ␤-stub) arising from ␣-secretase or ␤-secretase cleavage of FL-APP ( Fig.  2A, left panel). A similar result was observed in microsomal membrane fractions generated from fibroblasts of PS1 wildtype mice. In contrast, microsomal membrane fractions from fibroblasts of PS1 Ϫ/Ϫ mice or from HEK293 cells stably expressing the loss-of-function PS1-D385A mutant all showed significant inhibition in the production of the 6-kDa C-terminal fragment of APP (Fig. 2B). Similarly, there was significant inhibition in the production of the 6-kDa C-terminal fragment of APP when microsomal membranes were extracted from HEK293 cell lines expressing APP Swedish , and the resulting microsomal membranes were themselves then incubated with Compound E (a well documented ␥-secretase inhibitor of A␤ production (24)) ( Fig. 2B). When detergent-solubilized, immuno-purified, high molecular weight presenilin complexes (which contained C99-APP, PS1-NTF, PS1-CTF, and nicastrin) were investigated under the same assay conditions, the same ϳ6-kDa APP C-terminal band was also produced ( Fig. 2A, right  panel). The generation of the ϳ6-kDa fragment was highly sensitive to the detergent used. Thus, CHAPSO permitted robust production of this fragment, whereas other detergents, such as digitonin and dodecylmaltoside, supported this cleavage event less efficiently (data not shown). Triton X-100 and Nonidet P-40, both of which disassemble PS1 complexes, resulted in no C-terminal APP stubs or A␤ generation (data not shown). These data, when taken together, imply that the ϳ6-kDa fragment is an authentic, presenilin-dependent, proteolytic cleavage product.
PS1 FAD Mutations Enhance A␤ 42 but Inhibit ⑀-Stubs-To explore the effects of PS1 mutations on APP and Notch processing, we compared the production of A␤, APP ⑀-stubs, and NICD fragments in HEK293 cells stably expressing either wild-type PS1 or one of three different PS1 missense mutations associated with FAD (i.e. G206A, ⌬Exon 9, or L392V). Analysis of conditioned media from cultured cell lines (Fig. 4A, Table I). In agreement with prior data, A␤ 40 levels were not statistically increased in the PS1 mutant cells either in conditioned medium (PS1-wt: 1.0 Ϯ 0.11; PS1-L392V: 1.05 Ϯ 0.26; PS1-G206A: 1.17 Ϯ 0.07, p ϭ not significant) or in the cell-free preparations (PS1-wt: 1.09 Ϯ 0.03; PS1-L392V: 1.15 Ϯ 0.11; PS1-G206A: 1.22 Ϯ 0.13, p ϭ not significant). The intuitive expectation was that the increased ␥-secretase activity induced by these mutations would be accompanied by increased levels of both the N-terminal (A␤) and the C-terminal (⑀-stubs) cleavage products. In fact, microsomal membrane preparations containing mutant PS1 produced lower amounts of the ϳ6-kDa ⑀-stubs, regardless of whether the production of ⑀-stubs was assessed in the microsomal membranes themselves (Fig. 4B and Table I) or in the supernatant when the membranes were pelleted (not shown). These differences in ⑀-cleavage in microsomal membrane preparations from cell lines expressing mutant PS1 were statistically significant when compared with similar preparations from mocktransfected or PS1 wild type-transfected controls (PS1-L392V ϭ 0.54 Ϯ 0.07, p ϭ 0.05; PS1-G206A ϭ 0.57 Ϯ 0.12, p Ͻ 0.05; PS1 ⌬Exon9 ϭ 0.47 Ϯ 0.07, p Ͻ 0.05). Furthermore, despite increased production of A␤ 42 , MALDI-MS analysis of the C-terminal APP fragments revealed only the presence of ⑀-stubs beginning at positions 49, 50, or 52 (Fig. 3, right  panel). This decrease in the generation of ⑀-stubs was not due to differences in the expression of full-length APP, in the maturation or processing of ␤APP, or in the steady state levels of ␣and ␤-secretase generated C-terminal stubs of APP (which are the direct substrates for ␥-secretase and ⑀-secretase cleavage).

FIG. 2. Generation of C-terminal APP fragments. A, C-terminal (⑀-stub) generation from microsomal membranes (left panel) or purified PS1 complexes (right panel).
Microsomal membranes from stable HEK293 cells expressing APP Swedish (APP Swe ) or purified PS1 complexes from stable HEK293 cells expressing C99-APP were incubated at 37°C for 30 min or 2 h. An ϳ6-kDa ⑀-stub migrating below the major APP-CTFs (␣-stub and ␤-stub) generated from ␣and ␤-cleavage ⑀-stub was detected by Western blot with anti-APP-CTF antibody. (The variation in ␤-stub density reflects differences in ␤-stub abundance in cells used to prepare the PS1 complexes, which overexpressed C99-APP.) As shown in B, ⑀-cleavage is dependent on the presence of functional presenilin. ⑀-stubs were inhibited in microsomal membranes from fibroblasts of PS1 Ϫ/Ϫ mice, D385A mutant PS1-transfected HEK293 cells, and microsomal membrane preparations treated with ␥-secretase activity inhibitor Compound E.

PS1 FAD Mutations Reduce the Generation of NICD-
The site of ⑀-cleavage within the transmembrane domain of APP is similar to the site of S3-cleavage of Notch. However, studies using isocoumarin inhibitors (25) and the analysis of the effects of specific PS1 mutations have revealed that the biological functions of PS proteins in the generation of A␤ and NICD can be separated (13,26). To compare S3-cleavage of Notch with ␥and ⑀-cleavage of APP, we next investigated the effects of these FAD mutations on S3-cleavage of Notch. The same stable HEK293 cell lines expressing either wild-type or mutant PS1 were transiently transfected with Notch⌬E (16 -18). Notch⌬E encodes membrane-tethered Notch tagged at the C terminus with c-myc and is a substrate for presenilin-dependent cleavage at the S3-site (16). NICD formation was then followed in pulsechase experiments. In contrast to cells expressing wild-type PS1, those expressing mutant PS1 showed significant reductions in NICD formation to about half of the amount in wildtype cells (PS1-L392V: 0.59 Ϯ 0.06, p Ͻ 0.05; PS1-G206A: 0.48 Ϯ 0.11, p Ͻ 0.05; PS1 ⌬Exon 9: 0.56 Ϯ 0.09, p Ͻ 0.05) ( Fig.  5 and Table I).
C99-APP Mutations That Alter A␤ 42 Production Do Not Alter ⑀-Stub Generation-The above data demonstrated that, although FAD mutations in PS1 increase A␤ 42 secretion, they inhibit ⑀-stub generation. To determine whether mutations in the substrate for ␥and ⑀-cleavage (i.e. C99/C83-APP) might also have the same effect, HEK293 cells were transfected with the wild-type C99-APP, I45F mutant C99-APP (which increases A␤ 42 (19)), or V50F mutant C99-APP (which decreases A␤ 42 (19)). Equivalent amounts of total membrane from each cell line were incubated in the assay system as described above. In agreement with similar results published during the preparation of this manuscript (1), the amounts of ⑀-stubs generated by these cell lines were essentially equivalent, suggesting that C99 mutations do not affect the generation of ⑀-stubs (Fig.  6), although they significantly alter the production of A␤ 42 . DISCUSSION Our data reveal that all three of the pathogenic PS1 mutations tested here differentially affect several presenilindependent activities. These three mutations (each in different domains of PS1) cause an increase in the activity of ␥-secretase producing A␤ 42 , whereas simultaneously inhibiting both ⑀-cleavage of APP and S3-cleavage of Notch. This differential effect of these PS1 mutations provides insight into the complex, regulated, intramembranous proteolysis of type 1 membrane proteins such as APP and Notch. First, our data support the growing body of evidence that suggests that ⑀-cleavage of APP  and S3-cleavage of Notch are likely to be similar, if not identical, proteolytic processing events (1)(2)(3)(4)(5). The single difference that has been noted to date between the ⑀and S3-cleavages is that S3-cleavage of Notch is sequence-dependent, whereas ⑀-cleavage of APP is not (1)(2)(3)(4)(5).
Second, our data show that the increase in ␥ 42 -cleavage of APP is accompanied by decreases in ⑀and S3-cleavage activities. This conclusion is entirely compatible with prior data that suggest that FAD-pathogenic mutations have a loss-of-function effect on S3-cleavage of Notch when examined both in vivo in C. elegans and biochemically in vitro (7,11,12). The reduction in S3-cleavage of Notch is modest, which provides an explanation for why FAD-associated mutations in PS1 do not have clinical evidence of embryonic defects in humans, who are usually heterozygous for PS1 mutations, and for why transgenes encoding FAD-linked PS1 mutants are able to rescue the Notch phenotype of PS1 Ϫ/Ϫ mice (27). (We are, however, aware of one case with a compound heterozygote mutation in PS1 (M146V/S365T) in which the affected proband did have several minor congenital malformations (28)). Our data are also in accord with previously published data indicating that when the C. elegans sel-12 C60S loss-of-function mutant is expressed in mammalian cells as the human PS1-C92S homologue, it reduced processing of Notch and increased A␤ 42 production (13,14).
Finally, the reciprocal relationship between ␥ 42 -secretase and the ⑀-secretase/S3-cleavage suggests that the molecular events leading to the production of A␤ peptide are probably more complex than a single proteolytic cleavage. It seems likely that both the ␥-secretase and the ⑀-/S3-cleavage functions are presenilin-dependent. However, it remains unclear whether both ⑀and ␥-secretase cleavages are necessary for the generation of A␤, and if so, in which order these cleavage events occur. The fact that an authentic ␥-stub (containing residues 43-99) has never been observed has been taken to support the view that ⑀-cleavage may be the initial and necessary cleavage followed by a carboxyl-exopeptidase degradation back to residue 42. However, our data suggest that these two activities are more likely to be separate events. Indeed, the most parsimonious explanation for our data is that there are two pathways: a physiological pathway mediating ⑀and S3-cleavage (to generate signal transduction molecules such as AICD and NICD) and a default pathway that generates A␤ 42 . These functionally distinct pathways, however, could still be subserved by the same catalytic machinery. Thus, in the absence of the presenilin complex, neither pathway would be active. However, functional modulation of the presenilin complex (for instance, by missense mutations or by post-translational modification of the presenilins themselves) or of other components of the presenilin complex (e.g. glycosylation of nicastrin) might alter the relative balance of ␥and ⑀-cleavage pathways. We cannot, of course, preclude the alternate possibility that distinct structural components are involved in these two pathways. Our data, in agreement with previously published data, show that although many missense mutations in the substrate for ␥-/⑀secretase cleavage (C99-APP) can alter A␤ production, they do not modulate ⑀-cleavage (1,29). Such observations are compatible with the hypotheses that 1) the molecular mechanisms of substrate:catalysis interactions for A␤ generation and ⑀-stub/ NICD production are different and 2) PS1 mutations might primarily affect a regulatory function rather than directly execute catalytic cleavage.
Cumulatively, our data provide the basis for arguing that ␥-secretase cleavage, which results in the generation of A␤ 42 , is functionally distinct from the cleavages necessary for the production of signaling molecules such as AICD and NICD. More work must be undertaken to understand the molecular mechanisms underlying these distinct pathways. However, when taken in conjunction with other results, our data raise the possibility of generating compounds that can selectively inhibit FIG. 5. PS1 mutations inhibit the generation of NICD. HEK293 cell lines stably expressing wild-type PS1 or PS1 mutations (which were proven to enhance A␤ 42 production) were transiently transfected with Notch⌬E cDNA. NICD formation was then assessed by pulse-chase experiments with incubation for a 1-h incubation at 37°C. NICD generation in cells expressing PS1 mutations was significantly inhibited. n ϭ three independent replications.

TABLE I
Effects of FAD-associated PS1 mutations on production of A␤, -stub, and NICD Production of -stub, NICD and A␤ in vitro were analyzed from stably expressing PS1 wt, PS FAD (L392V, G206A, and ⌬Exon 9) mutations as described under "Experimental Procedures." Means Ϯ S.E. represent the results of more than three independent experiments. Ratios of -stub to CTF-APP or FL-APP and NICD to total Notch protein (NICD ϩ Notch DE) and values of A␤ level were normalized relative to those in cells expressing wild-type PS1, which was set at 1.00. p values indicated the significance (two-tailed Student's t test) relative to wild-type PS1. ND, not detectable. NS, not significant.